enhanced four-wave mixing in graphene-silicon slow-light ...€¦ · layer graphene.22,23 the...
TRANSCRIPT
![Page 1: Enhanced four-wave mixing in graphene-silicon slow-light ...€¦ · layer graphene.22,23 The transmission of the chip is tested by a tunable laser (Ando AQ4321A) scanning from 1510nm](https://reader036.vdocuments.mx/reader036/viewer/2022081406/5f170a0f663073087048d68b/html5/thumbnails/1.jpg)
Enhanced four-wave mixing in graphene-silicon slow-light photonic crystal waveguidesHao Zhou, Tingyi Gu, James F. McMillan, Nicholas Petrone, Arend van der Zande, James C. Hone, Mingbin Yu,
Guoqiang Lo, Dim-Lee Kwong, Guoying Feng, Shouhuan Zhou, and Chee Wei Wong
Citation: Applied Physics Letters 105, 091111 (2014); doi: 10.1063/1.4894830 View online: http://dx.doi.org/10.1063/1.4894830 View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/105/9?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Optimizing terahertz surface plasmons of a monolayer graphene and a graphene parallel plate waveguide usingone-dimensional photonic crystal J. Appl. Phys. 114, 033102 (2013); 10.1063/1.4813415 Polarization-independent slow light in annular photonic crystals Appl. Phys. Lett. 102, 141112 (2013); 10.1063/1.4801977 Wideband group velocity independent coupling into slow light silicon photonic crystal waveguide Appl. Phys. Lett. 97, 183302 (2010); 10.1063/1.3513814 Electro-optic polymer infiltrated silicon photonic crystal slot waveguide modulator with 23 dB slow lightenhancement Appl. Phys. Lett. 97, 093304 (2010); 10.1063/1.3486225 Solution-processed cavity and slow-light quantum electrodynamics in near-infrared silicon photonic crystals Appl. Phys. Lett. 95, 131112 (2009); 10.1063/1.3238555
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.59.150.244 On: Sat, 27 Sep 2014 16:54:53
![Page 2: Enhanced four-wave mixing in graphene-silicon slow-light ...€¦ · layer graphene.22,23 The transmission of the chip is tested by a tunable laser (Ando AQ4321A) scanning from 1510nm](https://reader036.vdocuments.mx/reader036/viewer/2022081406/5f170a0f663073087048d68b/html5/thumbnails/2.jpg)
Enhanced four-wave mixing in graphene-silicon slow-light photonic crystalwaveguides
Hao Zhou,1,2,a) Tingyi Gu,2,a) James F. McMillan,2 Nicholas Petrone,3
Arend van der Zande,3 James C. Hone,3 Mingbin Yu,4 Guoqiang Lo,4
Dim-Lee Kwong,4 Guoying Feng,1 Shouhuan Zhou,1,5 and Chee Wei Wong2,a)
1College of Electronic Information, Sichuan University, Chengdu 610064, China2Optical Nanostructures Laboratory, Columbia University, New York, New York 10027, USA3Mechanical Engineering, Columbia University, New York, New York 10027, USA4The Institute of Microelectronics, Singapore 117685, Singapore5North China Research Institute of Electro-Optics, Beijing 100015, China
(Received 18 June 2014; accepted 25 August 2014; published online 5 September 2014)
We demonstrate the enhanced four-wave mixing of monolayer graphene on slow-light silicon
photonic crystal waveguides. 200-lm interaction length, a four-wave mixing conversion efficiency
of �23 dB is achieved in the graphene-silicon slow-light hybrid, with an enhanced 3-dB conversion
bandwidth of about 17 nm. Our measurements match well with nonlinear coupled-mode theory
simulations based on the measured waveguide dispersion, and provide an effective way for
all-optical signal processing in chip-scale integrated optics. VC 2014 AIP Publishing LLC.
[http://dx.doi.org/10.1063/1.4894830]
Recently, graphene, with its unique dispersionless
and broadband band structure, provides an alternative plat-
form for condensed matter physics and nanoscale devices
including electronic, thermal, mechanical, and photonic
applications. In the optical domain, exciting graphene
enhanced applications include ultrafast mode-locking, mod-
ulators, detectors, photovoltaics, and optical frequency con-
version.1–6 Frequency conversion based on four-wave
mixing (FWM) in mesoscopic devices has been examined in
silicon nanowires and photonic crystal waveguides
(PhCWGs).7–10 In the communications (C) band, the bulk
Kerr coefficient n2 of silicon (�4� 10�18 m2/W) can achieve
enhanced four-wave mixing with engineered phase-matching,
long waveguides, and/or slow group velocity interactions in
engineered PhCWGs.11–15 Graphene has even larger third-
order nonlinearity susceptibility, a few orders of magnitude
larger than bulk silicon and other semiconductors, with a cor-
responding Kerr coefficient �10�13 m2/W. FWM is observed
in graphene recently,4,16–18 such as with microcavities and
detuning within the cavity linewidth.4 With the wider band-
width waveguide implementation, here, we show the experi-
mental observations of FWM in graphene-photonic crystal
waveguides. Our measurements demonstrated up to a 17 nm
(3-dB) conversion bandwidth in the slow-light region and 2D
efficiency maps of the graphene-silicon idler conversion with
a maximum �23 dB conversion efficiency, matching well
with our theoretical modeling.
Inset of Fig. 1(a) shows the scanning electron micro-
graph (SEM) of the air-bridged silicon photonic crystal
nanomembrane of 250 nm thickness and with monolayer gra-
phene transferred on the top surface. The PhCWG consists
of a missing row of holes in a W1 configuration with disper-
sion examined earlier,19 fabricated by deep-ultraviolet opti-
cal lithography and with an effective cross-sectional model
area (Aeff) about 0.55 lm2. In our measurements here, the
PhCWG total length is 200 lm. The monolayer graphene is
grown by chemical vapor deposition (CVD) and wet-
transferred onto the W1 PhCWG.20,21 The central PhCWG
area is covered by large-area monolayer graphene for
evanescent-field graphene interaction. Raman spectroscopy is
used to identify the single-layer graphene, with five selected
Raman spectra shown in Fig. 1(a). Excited by a 532 nm wave-
length laser, most of the graphene area has a G peak full-
width half-maximum (FWHM) about 20 cm�1, a 2D peak
FWHM about 38 cm�1, and a 2D-to-G peak intensity ratio
(�2). This indicates the PhCWG is covered by uniform mono-
layer graphene.22,23 The transmission of the chip is tested by a
tunable laser (Ando AQ4321A) scanning from 1510 nm to
1555 nm, as shown in Fig. 1(b). The transmission before
(red solid line) and after (blue solid line) transferred with
graphene are presented. The group index of graphene-
PhCWG is measured by the phase-delay method (green
dots) and fitted by black line.12 This transferred graphene
layer has an intrinsic absorption of �0.05 dB/lm (or about
10 dB) for this sample, at 1510 nm and far from the slow-
light band edge.
Fig. 2(a) shows the experimental setup used to achieve
the degenerate FWM in the graphene-PhCWG. Two
continuous-wave tunable lasers are amplified by two erbium-
doped fiber amplifiers (EDFA; IPG EDA-3K-C, and Amonis
AEDFA-PA-35-B-FA), serving as the pump and signal sour-
ces, respectively. A 50:50 optical fiber coupler combines the
pump and signal prior to coupling into the graphene-PhCWG
sample with a lensed fiber coupler. The coupling loss is
about 4 dB, for a maximum pump power coupled into the
waveguide of about 17 dBm. Another lensed fiber is placed
after the waveguide to collect the output into an optical spec-
trum analyzer (OSA; Advantest Q8384). This all-fiber sys-
tem is very robust. Before we mount the chips, the fiber to
fiber transmission measurement is taken to observe the loop
transmission as the background transmission in order to
a)Electronic addresses: [email protected]; [email protected]; and
0003-6951/2014/105(9)/091111/4/$30.00 VC 2014 AIP Publishing LLC105, 091111-1
APPLIED PHYSICS LETTERS 105, 091111 (2014)
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.59.150.244 On: Sat, 27 Sep 2014 16:54:53
![Page 3: Enhanced four-wave mixing in graphene-silicon slow-light ...€¦ · layer graphene.22,23 The transmission of the chip is tested by a tunable laser (Ando AQ4321A) scanning from 1510nm](https://reader036.vdocuments.mx/reader036/viewer/2022081406/5f170a0f663073087048d68b/html5/thumbnails/3.jpg)
obtain the actual chip transmission. The transmission differ-
ence is 61 dB or less.
Fig. 2(b) shows the observed FWM results in the
graphene-PhCWG sample (red solid), compared with the mon-
olithic silicon-only PhCWG (blue dashed). To avoid the ampli-
fied pump and signal transmission submerging the output idler
laser spectrum, the pump and signal wavelength are kept at
500 pm detuning. In this example, the pump wavelength is at
�1540 nm. We consider the Gidler ¼ Poutputidler =Poutput
signal by taking
the ratio of idler output power and signal output power. This
definition can cancel differences caused by any residual fluctu-
ation of the coupling condition. To ensure the FWM is not gen-
erated in the auxiliary fibers, the optical spectra before the
graphene-PhCWG sample is checked by the OSA and there is
negligible idler generated. In addition, for our measurements,
we tune the signal laser to TM-polarization which provides
more evanescent light interaction with graphene, resulting
higher conversion efficiency. The conversion efficiency in
graphene-PhCWG is about �23 dB compared to the mono-
lithic silicon-only PhCWG control sample at about �31 dB.
The FWM conversion efficiency, Gidler, in the graphene-
PhCWG device is described by11,12
Gidler ¼Pidler
Psignal¼
c�graphene�siliconPp
gsin h gLð Þ
!2
e�a�L; (1)
where a* is group index dependent propagation loss of the idler,
L is the waveguide length, Pp is the effective average pump
power (modified by loss a* along the waveguide length L), and
c*graphene-silicon is the group index dependent nonlinear parame-
ter of the graphene-silicon hybrid, c*graphene-silicon¼ c(ng/n0)2,
c¼ 2pn2/Aeffk.12 Here, Aeff is a wavelength-dependent term,
with larger cross-section at the slow-light region.19,24 n2 is the
overall graphene-silicon Kerr coefficient weighted by the E-
fields,4,25 and is computed to be �7.7� 10�17 m2/W. Here, g is
the parametric gain, described as11,12
g ¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffic�graphene�siliconPp
� �2
� DKL þ DKNL
2
� �2s
; (2)
where DKL � ðDxÞ2b2ðkpumpÞ þ ðDxÞ4b4ðkpumpÞ=12 and
DKNL ¼ 2c�graphene�silicon�Pp is the linear and nonlinear phase
mismatching, respectively. Dx ¼ jxpump � xsignalj, the sec-
ond and fourth order dispersion parameters (b2(k) and b4(k))
can be numerically calculated from the experimental disper-
sion curve of the device.11 With the large c*graphene-silicon
(and n2) and high group index, when we pump at 1540.0 nm
and probe at 1540.5 nm, with a simulated FWM conversion
efficiency Gidler of ��40 dB matches the experimental result
in Fig. 3(a).
The complete experimental and modeled FWM conver-
sion efficiency maps are shown in Figs. 3(a) and 3(b).
FIG. 2. The schematic of the experimental setup and the output spectra of the FWM before and after graphene is transferred onto PhCWG. (a) The experimen-
tal setup schematic diagram for degenerate FWM in graphene-PhCWG; (b) FWM spectra in graphene-PhCWG (red solid) and the control silicon-only
PhCWG (blue dashed). The pump wavelength is at �1540 nm, and the signal detuning is about 500 pm. The FWM conversion efficiency in PhCWG and
graphene-PhCWG are about �31 dB and �23 dB, respectively.
FIG. 1. The Raman spectra, transmission spectra, and group index of the Graphene-PhCWG device. (a) Five selected Raman spectra of the graphene on
PhCWG; (b) Transmission spectra before (red solid line) and after (blue solid line) graphene transfer as well as the measured group index of graphene-
PhCWG. Green dots are the measured group index of the graphene-PhCWG and the black line shows the fitted group index. Inset in (a) is the SEM of the
graphene-PhCWG device. The central area is continuously covered by large-area monolayer graphene.
091111-2 Zhou et al. Appl. Phys. Lett. 105, 091111 (2014)
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.59.150.244 On: Sat, 27 Sep 2014 16:54:53
![Page 4: Enhanced four-wave mixing in graphene-silicon slow-light ...€¦ · layer graphene.22,23 The transmission of the chip is tested by a tunable laser (Ando AQ4321A) scanning from 1510nm](https://reader036.vdocuments.mx/reader036/viewer/2022081406/5f170a0f663073087048d68b/html5/thumbnails/4.jpg)
Different pump wavelengths, with their group indices de-
pendence, and signal detunings are illustrated. In the meas-
urements, the pump wavelength is tuned from 1534.5 nm to
1540 nm with 500 pm step-sizes and the signal is detuned
from the pump between �4 nm and þ4 nm with 200 pm
step-sizes. The signal-pump detuning is defined as (ksignal �kpump). The pump and signal powers are kept constant in our
measurements at �9 dBm. The input power is slightly
decreased to avoid the device damage caused by intensity
fluctuations when detuning the wavelength. The simulations
are modeled in the same optical spectra as the measurements.
Both experiments and simulations show that, for each pump
wavelength examined (between 1534.5 nm and 1540.0 nm),
the conversion efficiency increases as the signal wavelength
is red detuned. As noted in Eq. (1) and previous studies,11,12
the highest conversion efficiency region should be when the
signal is detuned for optimal phase matching region and
when the group velocity is slowed for increased light-matter
interaction. Our measurements and modeling results show an
asymmetry in the FWM conversion efficiency because the
Graphene-PhCWG chip has an increased propagation loss
for the signal or idler light at the mode cutoff (towards the
longer wavelengths). For a more clear comparison of the
FWM conversion efficiency difference between Graphene-
PhCWG and monolithic silicon PhCWG, the pump wave-
length is fixed at 1540 nm and the signal is detuned from
�3 nm to þ3 nm, as shown in Fig. 3(c). When the signal
detunes from �1 nm to þ1 nm, the FWM conversion effi-
ciency is 3 dB to 6 dB enhanced by graphene. For long
detuning wavelength, the conversion efficiency is increased
and for short detuning wavelength the conversion efficiency
is decreased equal to that in silicon because of the increased
loss towards longer wavelengths. Moving into the longer
wavelength slow-light region for the pump, the FWM con-
version efficiency is somewhat constant due to the tradeoff
between the slow-light enhancement and the higher gra-
phene loss in this region. The conversion efficiency of
FWM in graphene-PhCWG and PhCWG when the signal
detuning is fixed at �0.2 nm and the pump wavelength is
tuned from 1536 nm to 1540 nm, as illustrated in Fig. 3(d).
Each data point show that the FWM conversion efficiency
is about 4 dB enhanced by graphene compared to mono-
lithic silicon-only PhCWG. We note that, with our
continuous-wave pump-signal lasers and at our �17 dBm
coupled-in chip powers, the nonlinear absorption and non-
linear losses are not significant since, when we increased
the pump power, the conversion efficiency increased line-
arly instead of decreasing, as shown in Fig. 4(a). Simulated
results of conversion efficiency vs. waveguide length are
shown in the inset of Fig. 4(a). The black line shows the
FWM conversion efficiency in silicon, which becomes con-
stant when the waveguide length is over 600 lm. The rela-
tionships between conversion efficiency of FWM in
graphene-PhCWG and waveguide length are indicated by
the red curve (graphene absorption 0.05 dB/lm) and blue
curve (graphene absorption 0.06 dB/lm). The conversion
efficiency increases fast with waveguide length because of
the high n2. And then it drops very fast because the FWM is
limited by the high absorption from graphene when the
waveguide length increases. This result indicates the
graphene-silicon hybrid structure can achieve higher con-
version efficiency in a shorter device with the proper band
gap engineering to decrease the absorption in graphene.
Our further work will focus on this.
FIG. 3. The experimental and modeling results of the FWM conversion efficiency map and the FWM conversion efficiency difference in graphene-PhCWG
and PhCWG. (a) Experimental 2D conversion efficiency map of graphene-PhCWG as a function of pump and signal wavelength, with pump laser from
1534.5 nm to 1540 nm with 500 pm step-sizes. The signal detuning is from �4 nm to þ4 nm with 200 pm step-sizes; (b) Corresponding modeling results of the
FWM conversion efficiency, with similar pump and signal ranges as the measurements; (c) Different conversion efficiency of FWM in graphene-PhCWG and
PhCWG when the pump wavelength is fixed at 1540 nm and the signal is detuned from �3 nm to þ3 nm; (d) The conversion efficiency of FWM in graphene-
PhCWG and PhCWG when the signal detuning is fixed at �0.2 nm and the pump wavelength is tuned from 1536 nm to 1540 nm.
091111-3 Zhou et al. Appl. Phys. Lett. 105, 091111 (2014)
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.59.150.244 On: Sat, 27 Sep 2014 16:54:53
![Page 5: Enhanced four-wave mixing in graphene-silicon slow-light ...€¦ · layer graphene.22,23 The transmission of the chip is tested by a tunable laser (Ando AQ4321A) scanning from 1510nm](https://reader036.vdocuments.mx/reader036/viewer/2022081406/5f170a0f663073087048d68b/html5/thumbnails/5.jpg)
Next, we investigate the 3-dB detuning bandwidth for
different signal detunings and for different central pump
wavelengths, as shown in Fig. 4(b). We define the whole
detuning range over which the idler output power is within
3-dB (61.5 dB) of the idler output power measured when the
detuning range is þ500 pm. (We did not use a 3-dB from the
peak FWM conversion efficiency because the FWM conver-
sion efficiency continuously increases for longer wave-
lengths due to the increased signal and pump scattering loss
at the longer (slow-light) wavelengths.) When the signal is
detuned to a shorter wavelength than the pump (negative
detuning D and for smaller absolute values of D), there is a
sharp decrease in the output idler intensity due to the higher
loss in the signal intensity. This higher loss arises from both
increased scattering loss in the slow-light region due to mi-
nute disorder in the PhC lattice and increased graphene linear
absorption. When the signal is detuned to a longer wave-
length than the pump (positive detuning D and for larger
absolute values of D), the output idler intensity is somewhat
constant due to relatively uniform transmission of the idler
in the normal group velocity region—this gives rise to the
large FWM bandwidth of 17 nm (at 1534 nm pump). This
bandwidth is several times larger than in monolithic silicon
PhCWGs. Within this �17 nm bandwidth, there are local
spectral fluctuations due to nonlinear phase-matching.
In this work, we demonstrate the enhanced FWM gener-
ation in silicon photonic crystal waveguides with monolayer
graphene, in which a �23 dB conversion efficiency and a
17 nm 3-dB detuning bandwidth are observed. We experi-
mentally and numerically analyze the effects of graphene
enhanced nonlinear coefficients, linear losses, the graphene-
PhCWG group indices, and waveguide length on the conver-
sion efficiency. The FWM conversion efficiency can be fur-
ther improved through engineering and reducing the linear
absorption of graphene, for all-optical signal processing in
next-generation communication modules.
The authors acknowledge discussions with Junbo Yang,
and support from the ONR program on nonlinear dynamics
(N000141410041), the National Science Foundation IGERT
(DGE-1069240), and the Chinese Scholarship Council.
1P. Avouris, Nano Lett. 10, 4285 (2010).2F. Bonaccorso, Z. Sun, T. Hasan, and A. Ferrari, Nat. Photonics 4, 611
(2010).3S. J. Koester and M. Li, Appl. Phys. Lett. 100, 171107 (2012).4T. Gu, N. Petrone, J. McMillan, A. van der Zande, M. Yu, G. Q. Lo, D. L.
Kwong, J. Hone, and C. W. Wong, Nat. Photonics 6, 554 (2012).5Q. Bao, H. Zhang, Y. Wang, Z. Ni, Y. Yan, Z. X. Shen, K. P. Loh, and D.
Y. Tang, Adv. Funct. Mater. 19, 3077 (2009).6F. Xia, T. Mueller, Y.-M. Lin, A. Valdes-Garcia, and P. Avouris, Nat.
Nanotechnol. 4, 839 (2009).7H. Fukuda, K. Yamada, T. Shoji, M. Takahashi, T. Tsuchizawa, T.
Watanabe, J.-I. Takahashi, and S.-I. Itabashi, Opt. Express 13, 4629
(2005).8R. Salem, M. A. Foster, A. C. Turner, D. F. Geraghty, M. Lipson, and A.
L. Gaeta, Nat. Photonics 2, 35 (2007).9R. Espinola, J. Dadap, R. Osgood, Jr., S. McNab, and Y. Vlasov, Opt.
Express 13, 4341 (2005).10M. A. Foster, A. C. Turner, J. E. Sharping, B. S. Schmidt, M. Lipson, and
A. L. Gaeta, Nature 441, 960 (2006).11M. Ebnali-Heidari, C. Monat, C. Grillet, and M. Moravvej-Farshi, Opt.
Express 17, 18340 (2009).12J. F. McMillan, M. Yu, D. L. Kwong, and C. W. Wong, Opt. Express 18,
15484 (2010).13C. Monat, M. Ebnali-Heidari, C. Grillet, B. Corcoran, B. Eggleton, T. White,
L. O’Faolain, J. Li, and T. Krauss, Opt. Express 18, 22915 (2010).14J. Li, L. O’Faolain, I. H. Rey, and T. F. Krauss, Opt. Express 19, 4458
(2011).15M. Dinu, F. Quochi, and H. Garcia, Appl. Phys. Lett. 82, 2954 (2003).16B. Xu, A. Martinez, K. Fuse, and S. Yamashita, paper presented at the
CLEO: Science and Innovations (paper CMAA6), May 1–6, 2011.17E. Hendry, P. Hale, J. Moger, A. Savchenko, and S. Mikhailov, Phys. Rev.
Lett. 105, 97401 (2010).18Y. Wu, B. Yao, Y. Cheng, Y. Rao, Y. Gong, X. Zhou, B. Wu, and K. S.
Chiang, IEEE Photonic Tech L 26, 249 (2014), see http://ieeexplore.
ieee.org/xpl/articleDetails.jsp?arnumber=6671974.19C. J. Chen, C. A. Husko, I. Meric, K. L. Shepard, C. W. Wong, W. M.
Green, Y. A. Vlasov, and S. Assefa, Appl. Phys. Lett. 96(8), 081107
(2010).20N. Petrone, C. R. Dean, I. Meric, A. M. Van Der Zande, P. Y. Huang,
L. Wang, D. Muller, K. L. Shepard, and J. Hone, Nano Lett. 12, 2751
(2012).21X. Li, W. Cai, J. An, S. Kim, J. Nah, D. Yang, R. Piner, A. Velamakanni,
I. Jung, and E. Tutuc, Science 324, 1312 (2009).22W. Zhao, P. H. Tan, J. Liu, and A. C. Ferrari, J. Am. Chem. Soc. 133,
5941 (2011).23A. C. Ferrari, J. C. Meyer, V. Scardaci, C. Casiraghi, M. Lazzeri, F.
Mauri, S. Piscanec, D. Jiang, K. S. Novoselov, S. Roth, and A. K. Geim,
Phys. Rev. Lett. 97, 187401 (2006).24L. O’Faolain, S. A. Schulz, D. M. Beggs, T. P. White, M. Spasenovic, L.
Kuipers, F. Morichetti, A. Melloni, S. Mazoyer, and J.-P. Hugonin, Opt.
Express 18, 27627 (2010).25S. Afshar V and T. M. Monro, Opt. Express 17, 2298 (2009).
FIG. 4. FWM conversion efficiency varies with the pump power and the waveguide length as well as the idler power varies with signal detuning. (a)
Conversion efficiency as a function of pump power (coupled in-chip power) at pump wavelength of 1540 nm; (b) Output idler power as a function of signal
detuning at different pump wavelengths. The maximum detuning bandwidth is about 17 nm when the pump wavelength is at 1534 nm. Inset in (a) shows the
simulation result of conversion efficiency vs. waveguide length. The black line shows the FWM conversion efficiency in silicon. The relationship between con-
version efficiency of FWM in graphene-PhCWG and waveguide length are indicated by the red curve (graphene absorption 0.05 dB/lm) and blue curve (gra-
phene absorption 0.06 dB/lm).
091111-4 Zhou et al. Appl. Phys. Lett. 105, 091111 (2014)
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.59.150.244 On: Sat, 27 Sep 2014 16:54:53